xor eax, eax

Simple Face Recognition for Frontal Profile Faces

So we are nearing the end of it … the most important part of it – Face Recognition. Well by now the face is detected in a photograph but what bout recognising those faces in a series of images in the Visual Databasae. Half our work is already done What remains needs to involve Artificial Intelligence.
We apply many edge masks to all faces to get an Eigen Face.
Edge Mask is some what like this

1 1 1     1 1 0     1 0 1     0 1 1
0 0 0     0 0 0     0 0 0     0 0 0
0 0 0     0 0 0     0 0 0     0 0 0

1 0 0     1 0 0     1 0 0     0 0 0
1 0 0     1 0 0     0 0 0     1 0 0
1 0 0     0 0 0     1 0 0     1 0 0

and so on …

After applying these edge masks the resulting image is an Eigen Face.

This Eigen Face is then fed to the input units of an Back propagation Neural Network which is already trained with a database for determining a face from another face, n will come up with the name of the person the face is tagged with.

Thats all for today guys … Hopefully soon i ll come up with a working algo for all this. Until Then .. Good Bye !!!

Simple Face Detection

So far so good. So, Skin Recognition is done .. however there might be objects in the image which have the same skin color but arent faces.. what bout those .. so our detection isnt complete, we need to detect faces in those skin areas of the new image I.

To cope up with this we need to the image I on various scales at various points in order to locate a face in the whole image. Scanning is done in all directions – left to right, top to bottom, starting a width n height of the image n reducing it to the minimum size unit. Now this can be a lengthy process n is very processor hungry… it is suggested to scan only at center of all locations in the image I which has a non zero Sij.

Now how do we determine whether an area is a face or not. This is done by subtracting a face template F from the region of the raw image A. The face template is calculated by averaging the pixel values over a number of face example images. After doing this in all directions the resulting image is called an Eigen Face.

Eij = |Aij – Fij|

This Eigen Face is then fed to the input units of an Error Back propagation Neural Network which is already trained with a database for determining a face from a non-face.Yea things have suddenly started getting over us .. isnt it .. Well if you are still with me then onto to the next part of it … Yes the last n final and the most important part of it … Recognising Faces …
Determining whether it is you, me or someone else .. Riya's Face recognition does this on the fly when u manually tag some of the faces n click on to Automatic Recognition Feature … So lets get started as to how is that done n get our hands dirty, punching it out.

Skin Detection Algorithm

An image is scanned n stored in 80×80 pixel array, A (say).

This array is then passed through a RGB threshold function to get three intensity images in Red , Green n Blue.

Rij = Arij – Tr

Gij = Agij – Tg

Bij = Abij – Tb

Here Rij, Gij and Bij are the red, green and blue pixel intensity values at image co-ordinates (i,j)
Arij, Agij and Abij are the red green and blue pixel intensities of the raw image

Tr, Tg and Tb are threshold values for Red Blue n Green respectively.

If the resulting values of Rij, Gij or Bij are less than zero then they are set to zero.

Now a new image S is calculated from the three images calculated.

Sij = 2Rij – (Gij + Bij)

Here Sij represents the likelihood of a skin tone colour at coordinates (i,j).

Now a new image I is calculated based upon the skin tone likelihood and the original raw image. If the algo finds a skin tone then it copies the raw image pixel intensity
value to the new image.
If  ( Sij>0  and Prij + Pgij + Pbij < Tw ){

Iij = (Prij + Pgij + Pbij)  /  3

}
Here Tw is a threshold value used to eliminate bright white colours(gamma) from the image.
So now we have an image I in which on Skin recognised parts are visible … Not bad start at all .. Let go further … 

Ever thought of how face recognition works ??? Riya’s face recognition might have surprised many but needs lil more tweaking … neway tis a good start … Have you ever wondered how face recognition works … The benefit of devising AI based sytem which is able to accurately recognise and locate objects / faces can really make a mark in computer vision technology.
All this needs lots of memory for all neccesary computations in order to process visual data out of the constraints of special light conditions n heavy costly gadgets for high resolution pictures.

Face / Object recognition is comparitively easier for special treated images where no other objects occlude, overlap or interfere with faces / objects in images… or that the background is not cluttered … however in realistic circumstances such things are quite obvious n cant be escaped for a robust algo.

Considering Riya’s Face recognition engine which considers all real time environments like objects, scenes n backgrounds, the best way, according to me, to track down a face in the whole image would be Skin Recognition.This is what their uploader does before uploading all images from your hard drives.l wouldnt have get onto how all this is done but since i got a few mails demanding the same i thought why not give it a try.. Believe me tis really simple …

So finally today i got sometime n i started playing with Riya's face n text reacognition. i started tagging faces with names n lil details like gender, email, and age. Everything seems to work fine except that Riya can only identify faces which have somewat frontal orientation. Riya allows both manual n automatic recognition. Ajax kungfu rules the interface … You can drag n drop using mouse or you can use ur tab key on the keyboard whichever suits your taste .. COOL .. yet one shortcoming that i found in Riya's interface was that they have assumed all users will have their desktop customised for 1024×768 or more. Seems their designers have overlooked the possibility.

You can keep on tagging images or use ur auto recognition feature .. watever u do the Riya's recognition engine keeps on working in the background … learning. Yet there is a stage in AI where more learning can also detoriate the recognition.Riya guys need to build better and better algorithms for recognising the trillions of faces which are getting pumped into it as fast as the available bandwidth will allow.

Riya's Uploader can be lil annoying at times coz it takes a lotta time uploading however it also has the advantage of providing the user a liberty to loaf around n in the meantime it uploads the snaps for them.

Now Riya or Flickr ??… Well am not telling that it's over between me and Flickr – I still think that Flickr is a very beautiful application and it's got many orders of magnitude in interface n functionality than Riya. Flickr is more userfriendly n artistic. If they somehow connect their visual database it ll make our life lot easier. After all both are tools that's fun to use.

I havent checked the community of Riya but i can expect that it has grown very fast.Am just waiting for the day when these Visual Database communities or public forums like Keyhole community where all users can upload such photos, recognize them n be able to track people all round the globe as to where they are right now .. Exciting isnt it …

i hv started building my app to achieve something like taht … yep my app was able to recognise faces using eigen faces in frontal profile however it still needs to track down faces in images of any size using Stero face detection..Yep all this needs lil time which am running short of these days .. neways .. lets hope for the best …

Flickr , an amazing site for photo management n sharing it online…Flickr has an amazing interface … Ajax makes it so cool not refreshing the pages everytime it opens. Along with Picasa, a software from Google's Image Team it makes a deadly combination coz uploading is so easy and fast. Flickr has a very good interface n u can upload 1GB / month .. not bad at all …

I had seen Ajax app before (also coded some ajax) .. some were also showcased at Barcamp Delhi … n this app from Flickr wins the race.

Here is My Flickr page located.

A couple of weeks ago i stopped by Riya another site in same regard however this has an edge over flickr, Riya can tags photographs by face recognition .. exciting isnt it … i bet it is .. however i lately tried riya n their installer was too slow to install n as far as i know their uploader takes a lotta time uploading pics to cyberspace… but the idea of face recognition really appeals me since i have tasted a bit of face recog earlier using eigen faces .. tis really challenging. Good Stuff Riya guys … i have signed up for their account but havent installed their uploader yet. Will take another couple of weeks owing to the busy schedule i have.

However Riya's web app has again forced n tempted me to look n code for face recognition n i have started playing around with it. Hopefully i ll soon publish it on my webpage…

Content Specialist for E-learning.Well … today I was offered the post of Content Specialist. Whether to join in or not is what is plagueing me. What is Content Specialist ? What does that mean ?? Well .. It means that if you develop software – in particular E-Learning n you can’t find the information you need, or say you need more information on a particular topic, then lemme know may be i can help you.

My job will be to write, edit, produce, and maintain within established timeframes and standards of quality and accuracy electronic and/or printed documentation for clients and staff.I ll have to design and create user documentation using both web and printed formats, develop and implement standards and procedures for end-user documentation preparation and publication, assist with planning and scheduling, and participate in software release team efforts and in testing user interfaces.

I’m gonna be your interface to question the excellent team of program writers helping them set priorities, by suggesting content and giving them feedback, and planning an overall documentation strategy… am goin to be the one asking you, what information you need, and how you want it to be told to others or in public.

I’m going to be annoying everyone by poking around, asking questions, trying to tweak things, trying out new ideas and usually making a nuisance of myself n get away telling everyone

" I’m doing it on your behalf."

Interesting isn't it … yep it is.

Well on to the real taste of it .. it asks a lotta hardwork .. I need to master central concepts, tools of inquiry, and structure of the disciplines of communication, mantaining currency in knowledge of the content area and skills of the same. tis gonna be a tough job… understanding major concepts, making assumptions, debates, processes of inquiry, of how student's world can influence their understanding and ways of knowing that are central to all disciplines, preparing our technical team for the global student community. The challenge coerced needs to be won by recognising multiple perspectives, conveying to learners how knowledge is developed from the vantage point of the knower.

The whole process of how the application of learning is useful, creating enthusiasm for all n how it connects in day to day life.

Which 3d package to opt for ??? well thats the question u get run over a hundred times if u are bout to start something related to 3d better get armed n have heart of a warrior

Trying to fetch an answer to the question i hv decided to manually check all packages n come up with the pros n cons of each …

I have a long list to take care of but it sure is the best way to determine our needs. Finding a 3d artist is really a big pain n paying him wat he demands is bigger pain in the ass. Yep when u are planning to release ur own game it really gets hectic to ur work done from ur team. most of all when u dont know what are they upto. To find out i decided to get my hands dirty …

Packages which i have tried til date are

Blender
http://www.blender.org/

Maya
http://www.alias.com/eng/products-s…ple/index.shtml

3d Studio Max
http://www.discreet.com/3dsmax/try.html
Softimage XSI
http://www.softimage.com/products/EXP/HL2/

Lightwave

http://www.fxrsoft.com/default.asp?page=support/download_lightwave.html

n Packages i havent tried yet are..

Rhino 3D
www.rhino3d.com

Cinema 4D
http://www.maxon.net

Milkshape 3d
http://www.milkshape3d.com/

gmax
http://www.discreet.com/products/gmax/

Softimage XSI
http://www.softimage.com/products/exp/v3/

Houdini
http://www.sidefx.com/products/apprentice/index.html

Truespace
http://www.caligari.com/store/trueS…ate=STtrueSpace

Wings 3d
http://www.wings3d.com/

Silo
http://nevercenter.com/index.php/Downloads/504

Lately today my friend called up asking help to uninstall Nokia Pc Suite from his pc .. he has no internet facility so it was very difficult for me to explaining every bit of shit over the phone for an hour or so, to no luck at the end .. Out of interest i jus called up a Nokia guy who clamied that the software could be only uninstalled by Nokia guys .. Holy crap .. neways i couldnt look into the problem myself as i had no access to his so called infected pc nor the software. watever it was i suggest the name of the software should be changed from Nokia PC Suite to Knock-ya PC Suite.

So …Guys please follow the procedure to play safe

1. Read the manual before installing any crap on ur pc…

2. Scan for any viruses in the package. n create log file for incremental compare on the whole sytem before and after installing.

3. Keep a back up of system image so that u can restore it if something goes wrong.

Dont mess up or pay for ur mistakes .. Customer care/ technician guys are waiting with their mouths open for baited clients like u.. BEWARE !!!

This just got in … checked my mail n we got one more project we were negotiating past few days. What drives me crazy is the fact that we are already meeting two deadlines. But tis a pleasure to have more work n no time, than no work n free time.

So am ready for the new lot … Whoa … am coming. Lately today also finished modelling my 3d head.Now Texturing remains … I hope i ll finish it off this weekend when am free.So twas a good day altogether.